Data integrity systems for organ contours in radiation therapy planning

Area of Science:

• Medical Physics
• Radiotherapy
• Image Analysis

Background:

• Accurate contouring of anatomical structures is critical in radiotherapy planning.
• Data integrity of contoured anatomy impacts treatment efficacy and patient safety.
• Existing methods for contour verification may be time-consuming or lack comprehensive detection capabilities.

Purpose of the Study:

• To develop and evaluate data integrity models for contoured anatomy in radiotherapy workflows.
• To differentiate between real-time and retrospective analysis suitability for different model types.
• To assess the accuracy and efficiency of novel contour integrity models.

Main Methods:

• Developed two classes of contour integrity models: data-driven and contiguousness models.
• Data-driven models compare contours against a gross set from similar disease sites for regions like bladder, spinal cord, and rectum.
• Contiguousness models analyze contour geometry using Extent and Region Growing metrics across various regions of interest.

Main Results:

• Data-driven models identified 70-80% of suspicious contours for bladder, spinal cord, and rectum.
• Contiguousness models demonstrated higher accuracy, with the Region Growing submodel being the most effective.
• Region Growing achieved 100% detection of noncontiguous contours and produced zero false negatives in most regions.

Conclusions:

• Contiguousness models, especially Region Growing, are highly accurate for detecting contour errors in radiotherapy.
• Contiguousness models are suitable for real-time clinical use, while data-driven models are better for retrospective analysis.
• The developed models enhance data integrity in radiotherapy contouring, improving treatment planning and patient safety.